Real-World 3D Tracking Dataset

该数据集包含三种可变形物体（绳子、布料和玩具动物）的多视角图像，并与每个时间步对应的机器人动作同步。为了提供全面的俯视视角并满足实验所需的多视角同步数据，该数据集是通过四个相机收集而成的。数据集涵盖了8个对象实例，跨越了3种类型的可变形材料。其应用任务包括三维追踪、基于动作的视频预测以及基于模型的规划。

This dataset contains multi-view images of three types of deformable objects (rope, fabric, and stuffed animals), which are synchronized with robot actions corresponding to each time step. To provide comprehensive top-down views and meet the requirements for synchronized multi-view data needed in experiments, this dataset was collected using four cameras. The dataset covers 8 object instances across 3 categories of deformable materials. Its applicable downstream tasks include 3D tracking, action-based video prediction, and model-based planning.